Developing data efficient algorithms in artificial intelligence
- In the past few years there has been an enormous amount of progress in machine learning, and one of the biggest contributing factors, especially for deep learning, is the vast amount of data that we have been able to collect, due to digitization and the internet. Harder and more ambitious problems in general artificial intelligence that that will enable agents to learn on their own and to act autonomously in the environment remain largely open. Initial breakthroughs include training an agent to play a complicated board game, or training agent to drive a car demonstrate that these problems require a lot of data even more data, even more compute than ever before, and possibly more than what we currently have available. This motivates several algorithmic challenges, namely how do we design algorithms that make the best use of the data that is available, and how do we design algorithms that are empirically and theoretically effective on the kinds of data that we often see in practice, for example, data with temporal dependencies and data that follow distributions that are hard to describe. This thesis proposes and analyzes a few algorithmic solutions along this theme, which is an important step to more reliably deploying general artificial intelligence into society.
|Type of resource
|electronic resource; remote; computer; online resource
|1 online resource.
|Van Roy, Benjamin
|Degree committee member
|Van Roy, Benjamin
|Stanford University, Department of Management Science and Engineering
|Statement of responsibility
|Submitted to the Department of Management Science and Engineering.
|Thesis Ph.D. Stanford University 2021.
- © 2021 by Xian Wu
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...